diff --git a/src/diffusers/modular_pipelines/minimax_h3/decoders.py b/src/diffusers/modular_pipelines/minimax_h3/decoders.py
index 44e9b2034f11..ed2266a3d481 100644
--- a/src/diffusers/modular_pipelines/minimax_h3/decoders.py
+++ b/src/diffusers/modular_pipelines/minimax_h3/decoders.py
@@ -184,7 +184,7 @@ def __call__(self, components: MiniMaxH3ModularPipeline, state: PipelineState) -
         latents_std = torch.tensor(components.vae.config.latents_std, device=device).view(1, -1, 1, 1, 1)
         latents = block_state.latents * latents_std + latents_mean
 
-        with torch.autocast(device_type=device.type, dtype=torch.float16, enabled=device.type == "cuda"):
+        with torch.autocast(device_type=device.type, dtype=torch.float16, enabled=device.type != "cpu"):
             video = components.vae.decode(latents, return_dict=False)[0]
         pixel_mean = torch.tensor(components.pixel_mean, device=device).view(1, -1, 1, 1, 1)
         pixel_std = torch.tensor(components.pixel_std, device=device).view(1, -1, 1, 1, 1)
diff --git a/src/diffusers/pipelines/wan/pipeline_wan_animate.py b/src/diffusers/pipelines/wan/pipeline_wan_animate.py
index a923219a7550..c2416243f111 100644
--- a/src/diffusers/pipelines/wan/pipeline_wan_animate.py
+++ b/src/diffusers/pipelines/wan/pipeline_wan_animate.py
@@ -465,8 +465,10 @@ def get_i2v_mask(
         mask_len: int = 1,
         mask_pixel_values: torch.Tensor | None = None,
         dtype: torch.dtype | None = None,
-        device: str | torch.device = "cuda",
+        device: str | torch.device | None = None,
     ) -> torch.Tensor:
+        device = device or self._execution_device
+
         # mask_pixel_values shape (if supplied): [B, C = 1, T, latent_h, latent_w]
         if mask_pixel_values is None:
             mask_lat_size = torch.zeros(